pith. machine review for the scientific record. sign in

arxiv: 2605.02640 · v1 · submitted 2026-05-04 · 💻 cs.AI

Recognition: 2 theorem links

· Lean Theorem

Trustworthy AI Suffers from Invariance Conflicts and Causality is The Solution

Ruta Binkyte , Ivaxi Sheth , Zhijing Jin , Mohammad Havaei , Bernhard Sch\"olkopf , Mario Fritz
Authors on Pith no claims yet

Pith reviewed 2026-05-08 19:31 UTC · model grok-4.3

classification 💻 cs.AI
keywords trustworthy AIcausalityinvariancefairnessrobustnessprivacyexplainabilityfoundation models
0
0 comments X

The pith

Causality resolves conflicts in trustworthy AI by allowing selective invariance to different data changes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that objectives like fairness, robustness, privacy, and explainability cannot be satisfied together because each requires a different kind of invariance when the underlying data-generating process is altered. It re-frames these conflicts as incompatible invariance demands and claims that causal models let practitioners choose which invariances to enforce while dropping others, thereby softening the trade-offs. A reader would care because current methods often improve one property only by harming another or by reducing overall performance, and the paper shows how causal thinking supplies a principled way to negotiate those compromises for both standard models and large foundation models.

Core claim

The central claim is that trustworthy AI trade-offs are best understood as incompatible invariance requirements that appear under different changes to the data-generating process, and that a causal framework supplies the tools to achieve selective invariance, thereby understanding how the trade-offs arise and how they can be softened or resolved while preserving utility.

What carries the argument

Re-interpretation of each trustworthy AI objective as an invariance requirement under specific changes to the data-generating process, combined with causal models that enable selective rather than uniform invariance.

If this is right

  • Causal assumptions can be applied explicitly or implicitly in large-scale foundation models to manage multiple objectives at once.
  • Trade-offs between performance and trustworthiness become easier to diagnose once each objective is stated as an invariance condition.
  • Selective invariance offers a route to models that are simultaneously fairer and more robust without uniform accuracy loss.
  • The same causal lens applies to both classical machine learning and modern foundation models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Practitioners could audit existing foundation models for implicit causal invariances by testing performance under targeted distribution shifts.
  • Mapping each trustworthy objective to a precise set of interventions might allow automated tools to suggest which invariances to keep or drop.
  • High-stakes domains such as healthcare could serve as testbeds to measure whether causal selective invariance actually improves real-world decision quality.
  • Future scaling laws for foundation models might need to include invariance cost as a variable when predicting trustworthiness.

Load-bearing premise

That the core trustworthy AI objectives can be fully expressed as invariance requirements under changes to the data-generating process and that causal selective invariance can reduce conflicts without creating new incompatibilities or losing utility.

What would settle it

A concrete case in which a causal model that enforces selective invariance for fairness and robustness still shows the same performance drop or new conflict that non-causal models exhibit would falsify the claim.

Figures

Figures reproduced from arXiv: 2605.02640 by Bernhard Sch\"olkopf, Ivaxi Sheth, Mario Fritz, Mohammad Havaei, Ruta Binkyte, Zhijing Jin.

Figure 1
Figure 1. Figure 1: Causal resolution to trade-offs in trustworthy AI. See Appendix A for the details. recent theoretical results show that under certain conditions robustness implies causality (Richens & Everitt, 2024). While the connection between causality and privacy is still less explored, early work indicates that causal structure can improve privacy-utility trade-offs (Tschantz et al., 2020; Tople et al., 2020; Binkyte… view at source ↗
read the original abstract

As artificial intelligence (AI), including machine learning (ML) models and foundation models (FMs), is increasingly deployed in high-stakes domains, ensuring their trustworthiness has become a central challenge. However, the core trustworthy AI objectives, such as fairness, robustness, privacy, and explainability, are hard to achieve simultaneously, especially while preserving utility. This position paper argues that causality is necessary to understand and balance trade-offs in performance and multiple objectives of trustworthy AI. We ground our arguments in re-interpreting trustworthy AI trade-offs as incompatible invariance requirements under different changes to the data-generating process. We then illustrate that causality provides a unifying framework for understanding how trade-offs in trustworthy AI arise, and how they can be softened or resolved through selective invariance. This perspective applies to both classical ML models and large-scale FMs. Our paper discusses how causal assumptions may be applied explicitly or implicitly in modern large-scale systems. Finally, we outline open challenges and opportunities for using causality to build more trustworthy AI.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. This position paper claims that trade-offs among trustworthy AI objectives (fairness, robustness, privacy, explainability) while preserving utility arise because these objectives impose incompatible invariance requirements under distinct changes to the data-generating process. It argues that a causal framework is necessary to understand these conflicts and to resolve or soften them via selective invariance, with the perspective applying to both classical ML models and large-scale foundation models; the manuscript discusses implicit or explicit use of causal assumptions in modern systems and outlines open challenges.

Significance. If the reinterpretation of objectives as invariance conflicts holds and causality uniquely enables selective invariance without new incompatibilities, the paper could offer a unifying conceptual lens for trustworthy AI research, potentially guiding method development beyond ad-hoc regularizers. A strength is the explicit grounding in re-interpretation of trade-offs and the extension to foundation models, which provides a coherent narrative even without new theorems or experiments.

major comments (3)
  1. [Abstract] Abstract and the core argument: the necessity claim ('causality is necessary to understand and balance trade-offs') is not supported by a derivation or counterexample showing that the listed objectives map to formally incompatible invariances under distinct DGP changes, nor that non-causal statistical or optimization methods cannot implement comparable selective invariance; without this, the conclusion that causality is required remains an assertion rather than a demonstrated requirement.
  2. [Re-interpretation of trade-offs] The section re-interpreting trustworthy AI trade-offs as invariance requirements: the mapping (e.g., fairness to invariance under protected-attribute interventions, robustness to invariance under covariate shifts) is presented conceptually without explicit formal definitions or proofs of incompatibility, which risks circularity because the invariance framing appears constructed to align with the proposed causal resolution.
  3. [Causality as unifying framework] Discussion of selective invariance via causality: the manuscript does not address whether causal assumptions themselves introduce new trade-offs or utility losses when applied to large-scale foundation models, nor does it compare against existing non-causal techniques that achieve partial invariance (e.g., via adversarial training or distributionally robust optimization).
minor comments (2)
  1. [Abstract] The abstract and introduction use 'invariance requirements' without an initial formal definition or reference to standard causal invariance literature (e.g., invariant causal prediction), which could be clarified for readers outside the subfield.
  2. [Open challenges] No concrete examples or pseudocode are provided for how selective invariance would be implemented in a foundation model setting, which would help ground the conceptual claims.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our position paper. As the manuscript is conceptual rather than a formal theoretical contribution, we clarify our arguments and indicate revisions to address the concerns about support for the necessity claim, potential circularity, and comparisons with alternative methods.

read point-by-point responses

  1. Referee: [Abstract] Abstract and the core argument: the necessity claim ('causality is necessary to understand and balance trade-offs') is not supported by a derivation or counterexample showing that the listed objectives map to formally incompatible invariances under distinct DGP changes, nor that non-causal statistical or optimization methods cannot implement comparable selective invariance; without this, the conclusion that causality is required remains an assertion rather than a demonstrated requirement.

    Authors: As a position paper, the core contribution is a unifying conceptual lens rather than a formal derivation of necessity. We will add a new subsection with concrete illustrative examples (e.g., fairness requiring invariance to protected-attribute interventions while robustness requires invariance to covariate shifts that can conflict under the same model). These examples will demonstrate incompatibility without claiming a general proof. We will also note that while non-causal methods can achieve partial invariance, they typically lack an explicit mechanism for selective targeting across multiple distinct DGP changes, which is the perspective we advance; a fuller comparison will be included. revision: partial

  2. Referee: [Re-interpretation of trade-offs] The section re-interpreting trustworthy AI trade-offs as invariance requirements: the mapping (e.g., fairness to invariance under protected-attribute interventions, robustness to invariance under covariate shifts) is presented conceptually without explicit formal definitions or proofs of incompatibility, which risks circularity because the invariance framing appears constructed to align with the proposed causal resolution.

    Authors: The mappings are drawn from established literature on causal fairness (invariance to protected-attribute interventions) and invariant representation learning (invariance to covariate shifts). To address circularity concerns, we will insert explicit definitions of each objective's invariance requirement, supported by citations to independent prior work that frames these objectives in invariance terms without reference to our causal resolution. This will clarify that the alignment follows from the literature rather than being constructed for the paper. revision: yes

  3. Referee: [Causality as unifying framework] Discussion of selective invariance via causality: the manuscript does not address whether causal assumptions themselves introduce new trade-offs or utility losses when applied to large-scale foundation models, nor does it compare against existing non-causal techniques that achieve partial invariance (e.g., via adversarial training or distributionally robust optimization).

    Authors: We agree these points require expansion. We will add discussion of limitations of causal approaches for foundation models, including computational costs of structure learning and risks of misspecification that may reduce utility. We will also include a comparison paragraph contrasting causal selective invariance with non-causal methods such as adversarial training (which enforces invariance via optimization) and DRO (worst-case robustness), noting that these achieve useful partial results but do not explicitly manage conflicts across multiple distinct DGP changes. revision: yes

Circularity Check

0 steps flagged

No circularity: interpretive position paper with independent conceptual reframing

full rationale

The paper is a position paper that proposes re-interpreting trustworthy AI trade-offs as invariance conflicts under data-generating process changes, then argues causality enables selective invariance to address them. This is presented as a unifying perspective rather than a formal derivation or prediction. No equations, fitted parameters, or self-citation chains reduce the central claim to its own inputs by construction. The argument relies on the proposed reinterpretation as its starting point and does not claim to derive new results from prior self-citations or data fits in a circular manner. The derivation chain is self-contained as an argumentative framework without load-bearing reductions to tautology.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper rests on domain assumptions about the difficulty of simultaneous trustworthy AI objectives and introduces an ad-hoc reinterpretation of trade-offs as invariance conflicts; no free parameters or new entities are defined in the abstract.

axioms (2)
  • domain assumption The core trustworthy AI objectives such as fairness, robustness, privacy, and explainability are hard to achieve simultaneously while preserving utility.
    Explicitly stated as the central challenge in the abstract.
  • ad hoc to paper Trustworthy AI trade-offs can be re-interpreted as incompatible invariance requirements under different changes to the data-generating process.
    This is the grounding step used to motivate the causal solution.

pith-pipeline@v0.9.0 · 5487 in / 1482 out tokens · 87996 ms · 2026-05-08T19:31:45.659592+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

300 extracted references · 57 canonical work pages

  1. [1]

    2021 , note =

    European. 2021 , note =

    2021

  2. [2]

    arXiv preprint arXiv:2407.15012 , year=

    Transformative influence of llm and ai tools in student social media engagement: Analyzing personalization, communication efficiency, and collaborative learning , author=. arXiv preprint arXiv:2407.15012 , year=

    work page arXiv

  3. [3]

    LearnLM, Team and Modi, Abhinit and Veerubhotla, Aditya Srikanth and Rysbek, Aliya and Huber, Andrea and Wiltshire, Brett and Veprek, Brian and Gillick, Daniel and Kasenberg, Daniel and Ahmed, Derek and others , journal=. Learn

  4. [4]

    A Survey of Large Language Models in Finance (FinLLMs),

    A survey of large language models in finance (finllms) , author=. arXiv preprint arXiv:2402.02315 , year=

    work page arXiv

  5. [5]

    Nature Medicine , pages=

    Toward expert-level medical question answering with large language models , author=. Nature Medicine , pages=. 2025 , publisher=

    2025

  6. [6]

    Khalaf, H., Verdun, C

    Decision-making behavior evaluation framework for llms under uncertain context , author=. arXiv preprint arXiv:2406.05972 , year=

    work page arXiv

  7. [7]

    2020 , note =

    Singapore Model AI Governance Framework , author =. 2020 , note =

    2020

  8. [8]

    proceedings of the 21th ACM SIGKDD international conference on knowledge discovery and data mining , pages=

    Certifying and removing disparate impact , author=. proceedings of the 21th ACM SIGKDD international conference on knowledge discovery and data mining , pages=

  9. [9]

    arXiv preprint arXiv:2408.04585 , year=

    Towards resilient and efficient llms: A comparative study of efficiency, performance, and adversarial robustness , author=. arXiv preprint arXiv:2408.04585 , year=

    work page arXiv

  10. [10]

    Fair inference on outcomes , Volume =

    Nabi, Razieh and Shpitser, Ilya , Booktitle =. Fair inference on outcomes , Volume =

  11. [11]

    Causally Testing Gender Bias in LLMs:

    Yuen Chen and Vethavikashini Chithrra Raghuram and Justus Mattern and Rada Mihalcea and Zhijing Jin , year = 2024, booktitle =. Causally Testing Gender Bias in LLMs:

    2024

  12. [12]

    Identifiability of Path-Specific Effects , booktitle =

    Chen Avin and Ilya Shpitser and Judea Pearl , editor =. Identifiability of Path-Specific Effects , booktitle =. 2005 , url =

    2005

  13. [13]

    Submitted to The Thirteenth International Conference on Learning Representations , year=

    MultiTrust: Enhancing Safety and Trustworthiness of Large Language Models from Multiple Perspectives , author=. Submitted to The Thirteenth International Conference on Learning Representations , year=

  14. [14]

    arXiv preprint arXiv:2401.15585 , year=

    Evaluating gender bias in large language models via chain-of-thought prompting , author=. arXiv preprint arXiv:2401.15585 , year=

    work page arXiv

  15. [15]

    Advances in neural information processing systems , volume=

    Investigating gender bias in language models using causal mediation analysis , author=. Advances in neural information processing systems , volume=

  16. [16]

    Findings of the Association for Computational Linguistics: EMNLP 2023 , pages=

    A causal view of entity bias in (large) language models , author=. Findings of the Association for Computational Linguistics: EMNLP 2023 , pages=

    2023

  17. [17]

    Using t-closeness anonymity to control for non-discrimination , abstract =

    Ruggieri, Salvatore , year =. Using t-closeness anonymity to control for non-discrimination , abstract =

  18. [18]

    2011 , eprint=

    Fairness Through Awareness , author=. 2011 , eprint=

    2011

  19. [19]

    Three naive Bayes approaches for discrimination-free classification , volume =

    Calders, Toon and Verwer, Sicco , year =. Three naive Bayes approaches for discrimination-free classification , volume =. Data Min. Knowl. Discov. , doi =

  20. [20]

    Kusner and Joshua R

    Matt J. Kusner and Joshua R. Loftus and Chris Russell and Ricardo Silva , year=. 1703.06856 , archivePrefix=

    work page arXiv

  21. [21]

    Kilbertus, M

    Niki Kilbertus and Mateo Rojas-Carulla and Giambattista Parascandolo and Moritz Hardt and Dominik Janzing and Bernhard Schölkopf , year=. 1706.02744 , archivePrefix=

    work page arXiv

  22. [22]

    Advances in neural information processing systems , pages=

    Equality of opportunity in supervised learning , author=. Advances in neural information processing systems , pages=

  23. [23]

    2018 , eprint=

    A comparative study of fairness-enhancing interventions in machine learning , author=. 2018 , eprint=

    2018

  24. [24]

    Statistical Analysis and Data Mining: The ASA Data Science Journal , volume=

    The fairness-accuracy Pareto front , author=. Statistical Analysis and Data Mining: The ASA Data Science Journal , volume=. 2022 , publisher=

    2022

  25. [25]

    Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , pages=

    Understanding and improving fairness-accuracy trade-offs in multi-task learning , author=. Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , pages=

  26. [26]

    Criminal Justice and behavior , volume=

    Evaluating the predictive validity of the COMPAS risk and needs assessment system , author=. Criminal Justice and behavior , volume=. 2009 , publisher=

    2009

  27. [27]

    Science , volume=

    Dissecting racial bias in an algorithm used to manage the health of populations , author=. Science , volume=. 2019 , publisher=

    2019

  28. [28]

    Conference on fairness, accountability and transparency , pages=

    Gender shades: Intersectional accuracy disparities in commercial gender classification , author=. Conference on fairness, accountability and transparency , pages=. 2018 , organization=

    2018

  29. [29]

    Management science , volume=

    Algorithmic bias? An empirical study of apparent gender-based discrimination in the display of STEM career ads , author=. Management science , volume=. 2019 , publisher=

    2019

  30. [30]

    L., Barocas, S., Daum \'e , III, H., and Wallach, H

    Language (technology) is power: A critical survey of" bias" in nlp , author=. arXiv preprint arXiv:2005.14050 , year=

    work page arXiv 2005

  31. [31]

    IEEE Technology and Society Magazine , volume=

    Bias and discrimination in AI: a cross-disciplinary perspective , author=. IEEE Technology and Society Magazine , volume=. 2021 , publisher=

    2021

  32. [32]

    arXiv preprint arXiv:2306.05068 , year=

    Shedding light on underrepresentation and Sampling Bias in machine learning , author=. arXiv preprint arXiv:2306.05068 , year=

    work page arXiv

  33. [33]

    , author=

    Algorithmic Bias in Autonomous Systems. , author=. Ijcai , volume=

  34. [34]

    ACM SIGKDD Explorations Newsletter , volume=

    On the applicability of machine learning fairness notions , author=. ACM SIGKDD Explorations Newsletter , volume=. 2021 , publisher=

    2021

  35. [35]

    Makhlouf, Karima and Zhioua, Sami and Palamidessi, Catuscia , journal=

  36. [36]

    Big data's disparate impact , author=. Calif. L. Rev. , volume=. 2016 , publisher=

    2016

  37. [37]

    2020 , publisher=

    Practical fairness , author=. 2020 , publisher=

    2020

  38. [38]

    Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , pages=

    Explaining algorithmic fairness through fairness-aware causal path decomposition , author=. Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , pages=

  39. [39]

    Shidler JL Com

    De-identified data and third party data mining: The risk of re-identification of personal information , author=. Shidler JL Com. & Tech. , volume=. 2008 , publisher=

    2008

  40. [40]

    Proceedings of the IEEE Symposium on Security and Privacy (SP) , pages =

    Narayanan, Arvind and Shmatikov, Vitaly , title =. Proceedings of the IEEE Symposium on Security and Privacy (SP) , pages =. 2008 , organization =

    2008

  41. [41]

    2007 IEEE 23rd international conference on data engineering , pages=

    t-closeness: Privacy beyond k-anonymity and l-diversity , author=. 2007 IEEE 23rd international conference on data engineering , pages=. 2006 , organization=

    2007

  42. [42]

    International journal of uncertainty, fuzziness and knowledge-based systems , volume=

    k-anonymity: A model for protecting privacy , author=. International journal of uncertainty, fuzziness and knowledge-based systems , volume=. 2002 , publisher=

    2002

  43. [43]

    ACM Transactions on Knowledge Discovery from Data (TKDD) , volume=

    l-diversity: Privacy beyond k-anonymity , author=. ACM Transactions on Knowledge Discovery from Data (TKDD) , volume=. 2007 , publisher=

    2007

  44. [44]

    Collecting Telemetry Data Privately , year =

    Ding, Bolin and Kulkarni, Janardhan and Yekhanin, Sergey , booktitle =. Collecting Telemetry Data Privately , year =

  45. [45]

    Learning with Privacy at Scale , volume =

    Differential Privacy Team, Apple , booktitle =. Learning with Privacy at Scale , volume =

  46. [46]

    Proceedings of the 2014 ACM SIGSAC Conference on Computer and Communications Security , year =

    \'Ulfar Erlingsson and Vasyl Pihur and Aleksandra Korolova , title =. Proceedings of the 2014 ACM SIGSAC Conference on Computer and Communications Security , year =. doi:10.1145/2660267.2660348 , publisher =

    work page doi:10.1145/2660267.2660348 2014

  47. [47]

    and Nissim, Kobbi and Raskhodnikova, Sofya and Smith, Adam , booktitle =

    Kasiviswanathan, Shiva Prasad and Lee, Homin K. and Nissim, Kobbi and Raskhodnikova, Sofya and Smith, Adam , booktitle =. What Can We Learn Privately? , year=

  48. [48]

    URLhttps://doi.org/10.1109/FOCS.2013.53

    John C. Duchi and Michael I. Jordan and Martin J. Wainwright , title =. 2013. doi:10.1109/focs.2013.53 , year =

    work page doi:10.1109/focs.2013.53 2013

  49. [49]

    Discrete distribution estimation under local privacy , author=. Int. Conf. on Machine Learning , pages=. 2016 , organization=

    2016

  50. [50]

    and Couchot, Jean-Fran c ois and Gambs, S \'e bastien and Palamidessi, Catuscia and Zolfaghari, Majid

    Arcolezi, H \'e ber H. and Couchot, Jean-Fran c ois and Gambs, S \'e bastien and Palamidessi, Catuscia and Zolfaghari, Majid. Multi-Freq-LDPy: Multiple Frequency Estimation Under Local Differential Privacy in Python. Computer Security -- ESORICS 2022. 2022. doi:10.1007/978-3-031-17143-7_40

    work page doi:10.1007/978-3-031-17143-7_40 2022

  51. [51]

    Proceedings of the 2018 International Conference on Management of Data , pages=

    Privacy at scale: Local differential privacy in practice , author=. Proceedings of the 2018 International Conference on Management of Data , pages=

    2018

  52. [52]

    arXiv preprint arXiv:2212.01627 , year=

    Castell: Scalable Joint Probability Estimation of Multi-dimensional Data Randomized with Local Differential Privacy , author=. arXiv preprint arXiv:2212.01627 , year=

    work page arXiv

  53. [53]

    Multi-Dimensional Randomized Response , year=

    Domingo-Ferrer, Josep and Soria-Comas, Jordi , journal=. Multi-Dimensional Randomized Response , year=

  54. [54]

    Warner , title =

    Stanley L. Warner , title =. doi:10.1080/01621459.1965.10480775 , year =

    work page doi:10.1080/01621459.1965.10480775 1965

  55. [55]

    Improving the utility of locally differentially private protocols for longitudinal and multidimensional frequency estimates , journal =

    H. Improving the utility of locally differentially private protocols for longitudinal and multidimensional frequency estimates , journal =. doi:10.1016/j.dcan.2022.07.003 , year =

    work page doi:10.1016/j.dcan.2022.07.003 2022

  56. [56]

    arXiv preprint arXiv:2301.01616 , year=

    Locally Private Causal Inference , author=. arXiv preprint arXiv:2301.01616 , year=

    work page arXiv

  57. [57]

    Tianhao Wang and Jeremiah Blocki and Ninghui Li and Somesh Jha , title =. 26th. 2017 , isbn =

    2017

  58. [58]

    URLhttps://doi.org/ 10.1007/11681878_14

    Cynthia Dwork and Frank McSherry and Kobbi Nissim and Adam Smith , title =. doi:10.1007/11681878\_14 , year =

    work page doi:10.1007/11681878

  59. [59]

    Foundations and Trends

    The algorithmic foundations of differential privacy , author=. Foundations and Trends. 2014 , publisher=

    2014

  60. [60]

    Privacy: Theory meets Practice on the Map , year =

    Machanavajjhala, Ashwin and Kifer, Daniel and Abowd, John and Gehrke, Johannes and Vilhuber, Lars , date-added =. Privacy: Theory meets Practice on the Map , year =. doi:10.1109/ICDE.2008.4497436 , journal =

    work page doi:10.1109/icde.2008.4497436 2008

  61. [61]

    Narayanan and V

    Narayanan, Arvind and Shmatikov, Vitaly , date-added =. De-anonymizing Social Networks , year =. doi:10.1109/SP.2009.22 , journal =

    work page doi:10.1109/sp.2009.22 2009

  62. [62]

    Distance Makes the Types Grow Stronger A Calculus for Differential Privacy , volume =

    Reed, Jason and Pierce, Benjamin , date-added =. Distance Makes the Types Grow Stronger A Calculus for Differential Privacy , volume =. 2010 , bdsk-url-1 =. doi:10.1145/1932681.1863568 , journal =

    work page doi:10.1145/1932681.1863568 2010

  63. [63]

    Privacy Enhancing Technologies: 13th International Symposium, PETS 2013, Bloomington, IN, USA, July 10-12, 2013

    Broadening the scope of differential privacy using metrics , author=. Privacy Enhancing Technologies: 13th International Symposium, PETS 2013, Bloomington, IN, USA, July 10-12, 2013. Proceedings 13 , pages=. 2013 , organization=

    2013

  64. [64]

    Miguel E. Andr. Geo-indistinguishability: Differential privacy for location-based systems , booktitle =. doi:10.1145/2508859.2516735 , year =

    work page doi:10.1145/2508859.2516735

  65. [65]

    International conference on theory and applications of models of computation , pages=

    Differential privacy: A survey of results , author=. International conference on theory and applications of models of computation , pages=. 2008 , organization=

    2008

  66. [66]

    2016 , publisher=

    Causal inference in statistics: A primer , author=. 2016 , publisher=

    2016

  67. [67]

    arXiv preprint arXiv:2412.12169 , year=

    Regulation of Language Models With Interpretability Will Likely Result In A Performance Trade-Off , author=. arXiv preprint arXiv:2412.12169 , year=

    work page arXiv

  68. [68]

    2015 , publisher=

    Counterfactuals and causal inference , author=. 2015 , publisher=

    2015

  69. [69]

    2009 , publisher=

    Causality , author=. 2009 , publisher=

    2009

  70. [70]

    Aristotle on causality , author=

  71. [71]

    1998 , publisher=

    Causality and explanation , author=. 1998 , publisher=

    1998

  72. [72]

    2017 , publisher=

    Causality and modern science , author=. 2017 , publisher=

    2017

  73. [73]

    Journal of the American Statistical Association , volume=

    Causal inference using potential outcomes: Design, modeling, decisions , author=. Journal of the American Statistical Association , volume=. 2005 , publisher=

    2005

  74. [74]

    Journal of Agricultural Research , volume =

    Correlation and causation , author=. Journal of Agricultural Research , volume =

  75. [75]

    Proceedings of the Eleventh conference on Uncertainty in artificial intelligence , pages=

    Causal inference in the presence of latent variables and selection bias , author=. Proceedings of the Eleventh conference on Uncertainty in artificial intelligence , pages=

  76. [76]

    2012 , publisher=

    Causality: Statistical perspectives and applications , author=. 2012 , publisher=

    2012

  77. [77]

    Journal of the American statistical Association , volume=

    Statistics and causal inference , author=. Journal of the American statistical Association , volume=. 1986 , publisher=

    1986

  78. [78]

    Annual Review of Statistics and Its Application , volume=

    Statistical causality from a decision-theoretic perspective , author=. Annual Review of Statistics and Its Application , volume=. 2015 , publisher=

    2015

  79. [79]

    2012 , publisher=

    Structural equations, graphs and interventions , author=. 2012 , publisher=

    2012

  80. [80]

    American philosophical quarterly , volume=

    Causes and conditions , author=. American philosophical quarterly , volume=. 1965 , publisher=

    1965

Showing first 80 references.